Rapid Detection Ingredient And Identification Anthenticity Of Semi-Dried Purple Sweet Potato Noodles By Naer Infrared Spectroscopy

Due to its bright color, high nutritional value, easy cooking, potato fragrance and other characteristics, the semi-dried purple sweet potato noodles have been popular among consumers. With the increasing development of purple sweet potato products, the fake purple sweet potato noodle which was made from pure pigments, began to appear in the market. In addition, its high water content makes the total viable bacteria count (TVC) grow faster, storage period become shorter and makes it unable to determine instantly whether the content of TVC meets national safety standards. This paper aims to use near-infrared spectroscopy (NIR) technology to provide a rapid detection method, detect the nutritional ingredient and edibility of sample, and identify the authenticity of samples by using the qualitative discrimination.1.In order to improve the prediction ability and stability of the model, the optimal sampling mode and preprocessing methods of spectra were chosen. The results showed that the dry background, the resolution 16cm-1and the number of scanning32times were chosen as the optimal sampling way under the condition of the semi-dried purple sweet potato noodles sifted through50mesh sieve,0.6g sample loaded into the sample bottle and pressured to uniform firmness. The results were finally recorded as log (1/R). Compared with different spectral preprocessing methods, the first derivative (ID), the second derivative (2D), multiplicative scatter correction (MSC), the extended multiplicative scatter correction (EMSC) and standard normal variable transformation (SNV) spectral preprocessing methods were able to improve the accuracy of models.2. The evaluation criteria of model optimization was relied on the correlation coefficient of calibration set (Rc), correlation coefficient of prediction set (Rp), root mean square error of cross validation of calibration set (RMSECV), root mean square error of prediction set (RMSEP), PRESS value and principal components. Comparing with10kinds of spectral preprocessing methods, the optimal models for anthocyanin content, purple sweet potato powder, starch content and protein content were established by using partial least squares (PLS). The optimal PLS model of anthocyanin content was treated with SNV+2D, the Rc, Rp, RMSECV and RMSEP were0.9872,0.9598,0.125and0.220. The optimal PLS model of purple sweet potato powder was treated with1D and the Rc, Rp, RMSECV and RMSEP were0.9894,0.9882,1.316 and1.375. The optimal PLS model of starch content was treated with MSC and the Rx, Rp, RMSECV, RMSEP were0.9399,0.9067,2.826and4.498. The optimal PLS model of protein was established with MSC+2D pretreated, the Rc, Rp, RMSECV, RMSEP were0.9811,0.8934,0.050and0.116.3. The total viable bacteria count (TVC) was directly related to the edibility of semi-dried purple sweet potato noodles. Comparing the10kinds of spectral preprocessing methods, the optimal model for TVC was established by using partial least squares (PLS). The optimal PLS model of TVC was treated with SNV+2D and the Rc, Rp, RMSECV and RMSEP were0.9921,0.9754,0.251and0.445. This result indicated that the optimal PLS model has a better predictive ability, we can accurately obtain the TVC in semi-dried purple sweet potato noodles by this optimal model.4. The discriminant analysis method of principal component analysis combined with mahalanobis distance (PCA-DA) was used to identify six kinds of semi-dried purple sweet potato noodles under different storage time. The samples’storage time of Od,2d,4d,6d,8d,10d were chosen. Comparing the9kinds of spectral preprocessing methods, the performance of SNV for DA model was superior to other models. The misclassification number was0, the recognition rate and variability described achieved100%. 5. In this study, the UV-visible spectroscopy, electronic nose technology and NIR spectroscopy combined discriminant analysis were used to identify the authenticity of semi-dried purple sweet potato noodles. UV-visible spectroscopy was based on the number and location of characteristic peaks to distinguish the authenticity of samples. In this study, the true sample has one characteristic absorption peak at530nm and the fake sample has two characteristic absorption peaks at520nm and630nm. The true sample has special flavor of sweet potato while the fake does not contain. Useing the electronic nose technology, true and fake samples could be distinguished effectively by using principle component analysis (PCA) and linear discriminant analysis (LDA). In the PCA discriminant analysis, the discrimination power between true and fake sample was0.789. The NIR spectroscopy soft independent modeling of class analogy (SIMCA) discriminant analysis method was used to identify the authenticity of samples. After the comparison of different spectral preprocessing methods, it finally obtained the optimal SIMCA model with SNV preprocessing, with the misclassification number was0and the correct recognition rate achieved to100%.